On 05/26/2015 08:16 PM, Paolo Bonzini wrote:
On 26/05/2015 12:15, Alexey Kardashevskiy wrote:
There was a "[RFC PATCH 00/15] spapr: add support for PHB hotplug"
patchset from Mike, this patch added "unrealize" for spapr_phb:
[RFC PATCH 05/15] spapr_pci: add PHB unrealize
I believe I am dealing with the fixed version of this patch so I'll ask
Mike to respin it.
So, the sentences after that one note an exception for alias and
container regions. I think iommu regions should behave similarly
- in a sense they're just a procedurally generated collection of
alias regions.
The difference is that containers and aliases are resolved at the time
the memory region tree is flattened, while IOMMU regions are resolved
at run time.
So they are not parts of flattened view and I should be able to
add/remove these IOMMU subregions any time I like?
Yes.
I got lost here:
>>> If it's not true now that they can be unparented at any time like
>>> alias regions, we should probably try to make it true.
>>
>> Unfortunately it's not so easy...
Uff. Tricky :)
memory_region_del_subregion() is not unparenting but just a wrapped
object_unref(), right? But since iommu MR are resolved dynamically, the
whole conversation we are having here now has nothing to do with my&Mike
concern what we can and cannot do with DMA windows here. Is this correct?
--
Alexey
